grad in model.named_parameters() are None

[asudeeaydin]

Hi,

I'm trying to plot the gradients of the network, but when I looped over the layers, I realized that grad was None. This made me a bit worried about the training altogether.

Maybe this is expected since a surrogate gradient function is used? In that case, is there no way to plot the gradients?

[fangwei123456]

Can you show me the minimul codes to reproduce your problem?

[asudeeaydin]

When trying to condense the code I found the mistake, it was due to the loss function.

Thanks for your quick response, really appreciate it!